Hacker Read top | best | new | newcomments | leaders | about | bookmarklet login

It's not a lot faster for input but it is something like 10x faster for output(mixtral vs gpt-3.5). This could enable completely new mode of interaction with LLMs e.g. agents.


sort by: page size:

The model is quicker to evaluate. So quicker responses and more throughput.

It's faster and/or easier to implement. Sometimes a lot

When used for business logic they also execute about 20x faster than the same logic encoded in a client, and in far fewer LOC. Getting rid of all those round-trips has a huge effect!

It's definitely not objectively better. For normal workflows, the startup latency is a real hindrance, even if the community has collectively built norms around that which help a bit (keeping a long running repl session open, etc.). But in terms of expressiveness plus attainable performance it really is hard to beat. I think it's nice that the path from quick draft to really performant code is continuous, and not a big gap like switching languages.

faster, it has options like 'modify'. I also feel it follows my commands better, esp. when i ask to rephrase

> it's significantly easier to work with

Can you please elaborate?


It’s far better at producing workable code, logical reasoning, keeping a train of thought, etc. Their benchmark breaks it down.

https://openai.com/research/gpt-4


Right. I understand those advantages. I was curious about the comparison of input parameters specifically.

If it's faster to generate? I don't know, that's what I am asking

It’s faster for coding.

Outright speed, speed of scalability.

If you can submit what you want and have it within moments, it beats a fancy typing pool. That kind of speed alone lets you have a highly trained someone retry different variations and filter more on quality of result.

Scalability. If you need to do it ten times faster. More hardware = done. More people could make it done too but training etc is required.


Massively higher potential output. It's the difference between sending spam e-mails out by hand and having a bot do it.

It is significantly faster, see benchmarks. Also more concise/ergonomic and declarative.

I think the most important gain is that client requests can fit in fewer packets -> fewer RTTs (critical on poor networks) -> lower latency to first paint -> better UX.

It's vastly faster and more reproducible, enough so as to make it a different environment with different considerations.

More parameters, more training time, faster processing speed to accomplish both of the first two.

Yes, faster load times. Also, a more sensible target language.

I find it more intuitive, and for a small number of iterations it's definitely faster.

It's faster (better performance). It could also be preferred just as a coding style.
next

Legal | privacy